Team D: AI and Data Economy

Mission: “Use data, Artificial Intelligence and innovation to transform the prevention, early diagnosis and treatment of chronic diseases by 2030”

The problem

Obesity and Greenspace

• The World Health Organization states that worldwide obesity has tripled since 1975.
• Health Survey for England found that of every 1,000 adults, 643 are classed as overweight to morbidly obese.
• Public Health England found that across adult males and females, spending less that 30 minutes per week engaged in moderate to vigorous physical activity is a key factor in obesity levels.

“Research suggests we need a strategic approach to the provision of parks and green spaces by identifying areas where investment will have the most significant impact on individuals.” A quote from Fields in trust who also estimated parks & green spaces are estimated to save the NHS £111 million/year.

• There is a variety of studies that finds a positive but weak association between obesity and availability of green spaces and sports facilities (Bedimo-Rung (2005); Lachowycz and Jones (2011); Lee and Maheswaran (2011); Huang et al. (2015); Jean-louis (2018))

How will we address this mission?

We aim to address this by…

• Investigating the spatial distribution of obesity in England
• Evaluating the importance of green spaces and sports facilities
• Assessing the significance of other potential explanatory variables

Our research question

Stakeholders:

1. Governmental:
   • Department of health and social care
   • Public Health england
   • NHS England and NHS Improvement
   • Department of Digital, Culture, Media and Sport
   • UK Sport, Sport England
2. Academics:
   • Academic Health Science Centre
3. Organisations:
   • Health Data Research UK
4. Healthcare users:
   • Patients, family, carers
5. Sport & health related policies:
   • Everybody active, every day

Datasets

We downloaded this data from public health england via fingertipsR

Data Availability

We used the finest scale available to us (Local authority district level) whilst also conforming to the NHS’s requirement of anonymity.

Analysis, theory and code

We used a general linearised geostatistical model originally developed in Besag, York, and Mollié (1991), and elaborated on in Brown (2015). It shares similarities with a Poisson regression model but properly harnesses spatial relationships. All code is available on slack/github.

Results

## Warning: namespace 'INLA' is not available and has been replaced ## by .GlobalEnv when processing object 'obesity_model'

Discussion

General Recommendation: A geographically-targeted whole system approach

Why is this a good idea?

• the effectiveness of individual measures to tackle obesity, at a population level, is limited
• a whole system approach is costly in a climate of limited resources
• geographically-targeted interventions can identify a) the areas most in need, and b) those likely to result in the greatest reward
• enables local authorities to take ownership of a problem affecting their district

What are the caveats?

• measures to evaluate the effectiveness of any intervention need to be in place to asses the model
• requires long-term investment and will be slow over time

Data recommendations & Future research:

• As granular-as-possible public health data, subject to practical and ethical considerations.
  • Point locations are ideal
  • Current trend in geospatial statistical modelling is how to work with such aggregated data as we have
• Ideal dataset has…
  • Well-defined outcome measures and covariates
  • Point locations, or as-granular-as-possible aggregation
• After identifying target areas, local investigation of underlying issues
• Assess economic viability of different socioeconomic issues (i.e. what is worth investing in?)

Ethics :

Although we identified the potential targeting areas that need interventions, the policy recommendation was made without the reinforcement of health inequalities. The further interventions using data-driven approach needs to protect the public’s right and privacy.